I've done my own tests with my body, lens combinations.
It IS true that with all lenses, auto-focusing isn't exact. But, except cases where there were mechanical problems, differences were very small.
For example, with my D200 and Nikkor 70-200 f/2.8VRII. On a tripod with remote release, shooting a test target, I can see the focusing variations at 100% magnifications. But the differences are so small, in more reasonable magnifications such as 16x20, it just doesn't matter. I mentioned this when I sent this lens in for some other thing and the note from tech says "within spec." I have other lenses in this class that perform very similarly.
Another example, (not Nikon) Tokina lens exhibited much bigger variations. Sent it in. Tech replaced the whole optical train and calibrated. Came back performing much better. In fact, there are virtually no variations
Yet another example, Nikkor 50mm f/1.8D. This was an interesting one. Ordered something like 3 to 4 of these and kept the best. Still did pretty badly wide open. Sent it in. Tech replaced the front element (this WAS a brand new lens!) and calibrated. Now it's good.
I also have all-plastic-consumer lenses. Many of them have plastic mounts. I actually have no complaints on focus accuracy or optical performance. Of course, them being not f/1.8 or 2.8, DOF is bigger so focusing errors affect less. Surprisingly, I had no mechanical problems either. They have proven to me they are quite durable despite pretty much everything being plastic. I fully expect these lenses won't last another 10 years, not less 30 years like some old pre-ai lenses. But I didn't pay that much money either. In old days, there were no such thing as consumer grade lenses. (except for E-series). Inflation adjusted, these old lenses were expensive.
Of course, in sub-par lighting conditions, accuracy do suffer. I have experienced this as well. But I wouldn't be able to do that by eyes much better either. Personally, I'm quite satisfied. Results are the same with my three F-100s as well.